Comment by ethangk

I'd love to know more about this. Is this all done 'offline', fine tuning one of the existing open models? Or is everything being sent to an OpenAI API?